[alsa-devel] [patch 1/1] sgtl5000: Fix suspend/resume

Arnaud Patard (Rtp) arnaud.patard at rtp-net.org
Tue Apr 5 09:56:12 CEST 2011

Mark Brown <broonie at opensource.wolfsonmicro.com> writes:

> On Mon, Apr 04, 2011 at 05:13:00PM +0200, Arnaud Patard wrote:
>>  /* default value of sgtl5000 registers except DAP */
>> -static const u16 sgtl5000_regs[SGTL5000_MAX_REG_OFFSET >> 1] =  {
>> +static const u16 sgtl5000_regs[SGTL5000_MAX_REG_OFFSET] =  {
>>  	0xa011, /* 0x0000, CHIP_ID. 11 stand for revison 17 */
>> +	0,
> It's not immediately obvious that this is the best fix - the driver does
> declare a step of 2 so I'd expect it to be able to lay out the cahce
> defaults like this.  Also, it'd be easier to review the patch if you

The sgtl5000 driver is using:

static const u16 sgtl5000_regs[SGTL5000_MAX_REG_OFFSET >> 1] =  {


.reg_cache_size = ARRAY_SIZE(sgtl5000_regs),

Which means that reg cache size is SGTL5000_MAX_REG_OFFSET / 2 and
register values stored without any "holes" in the array, except that in
flat cache case, ASoC does :

static int snd_soc_flat_cache_read(struct snd_soc_codec *codec,
                                   unsigned int reg, unsigned int
        *value = snd_soc_get_cache_val(codec->reg_cache, reg,
        return 0;

and snd_soc_get_cache_val returns cache[reg] which means that the cache
size must have a length of SGTL5000_MAX_REG_OFFSET and values stored
with "holes".

Also, without this change, the codec_reg debugfs files is returning
lines like :
0000: a0
instead of :
0000: a011

This is also the reason why resume was broken: the restored values were
not the right ones. Look at the restore regs code: instead of reading
cache[reg] like snd_soc_get_cache_val, it was reading cache[reg>>1].


More information about the Alsa-devel mailing list